PWV Portfolio Feed

✨2025 recap: what we shipped for agent workflows 👇✨

🤖 Agent Runners
🔌 MCP Server
🗄️ Netlify DB
🧠 AI Gateway
📈 Observability
🌅 Prerendering
🧷 Skew Protection

Links + why it matters in thread 🧵

AI Gateway is basically “LLM access as a Netlify primitive.” 

It proxies model calls from Netlify runtime so you don’t need to maintain provider accounts + API keys across environments.

Cost: On Netlify, AI inference is priced off the underlying model provider costs, then converted into Netlify credits ($1 USD of model usage = 180 credits).  
OpenRouter generally passes through model pricing and adds a 5.5% fee.

Good Q. Both Netlify AI Gateway and OpenRouter give you a single API surface to multiple models.

Main difference: Netlify AI Gateway is built into the deploy/runtime workflow on Netlify (so you can pair it with things like Functions/Edge + platform controls). OpenRouter is model-routing first.

Doing clean builds (with no binaries) in large Xcode codebases is so unproductive. I don't know how teams can work this way. 7580/13083 targets and still going...

One place where agents genuinely shine is navigating the AWS console. Instead of clicking through endless menus trying to find that one setting, just ask and let the agent figure out where everything lives.

Wishing you deterministic workflows in a probabilistic world.

Or at least fewer surprises. 💚

Next week: a quick “what we shipped in 2025” thread focused on agent workflows, shipping faster, and debugging less. 

Drop your favorite Netlify moment of the year 👇